JPH04342001A - Apparatus for controlling process variable and adaptive tuning method of adaptive controller - Google Patents

Abstract

PURPOSE: To automatically control a process in optimum by tuning the parameter of adaptive controllers. CONSTITUTION: The adaptive controllers 21 and 22 receive error signals and generate output control signals connected to a process 10. The adaptive controllers 21 and 22 are tuned by the process 10 so that the process 10 outputs process variable PV signals having a prescribed characteristic with the output control signals. Process variable PV signals are connected to the second input terminal of a differential device 23 and the adaptive controllers 21 and 22 and therefore the adaptive controllers 21 and 22 can independently monitor the process variable PV signals. The arbitrary change of the process variable PV signal as the result of the disturbance of a process parameter can be detected with such constitution and the tuning of the adaptive controllers 21 and 22 is adjusted. Thus, the process 10 can automatically be controlled in optimum.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a PID controller, and more particularly to an automatically tuned PID controller and an adaptive control method therefor.

0002

BACKGROUND OF THE INVENTION In conventional feedback process control devices, the deviation (e) between the set point (sp) and the controlled variable (pv) fed back from the process is calculated using proportional + integral +
A differential operation (PID operation) is performed, and the result of the PID operation is supplied to the process as a control signal. In order to optimally control the process, the PID parameters for performing each PID operation are set to optimal values. Traditionally, PID parameters were adjusted manually. Step response and limit sensitivity methods are well known for performing manual adjustments. However, in these two methods, it takes a long time to measure the characteristics, and the process control is stopped during the measurement, so the pv value obtained at that time cannot be the most desirable value.

On the other hand, methods have been proposed in which the PID parameters are determined automatically (rather than manually). Such an automatic method is described in US Pat. No. 4,754,391. However, if any changes occur in the process, the controller must be returned manually.

In the adaptive controller and method of the present invention, process output is monitored. Therefore, assuming that sufficient process changes occur after lineout (i.e., sp = pv) to cause a given error without causing a corresponding change in the set point (sp), the adaptive control of the present invention The instrument begins analyzing the process and retunes the PID parameters to optimize control of the process.

[0005]

SUMMARY OF THE INVENTION It is therefore an object of the invention to provide an adaptive controller. Another object of the invention is to obtain an adaptive controller that can be automatically retuned to provide optimal control of the controlled process. Another object of the invention is to obtain an adaptive controller that directly monitors the output of the controlled process. Another object of the invention is to provide an adaptive controller that directly monitors the output of a connected process and initiates retuning of the adaptive controller when it detects that a predetermined condition exists. Yet another object of the present invention is to provide an adaptive controller and an adaptive control method that automatically retune the adaptive controller when predetermined conditions exist in order to optimally control the controlled process.

[0006]

SUMMARY OF THE INVENTION The present invention provides an adaptive controller and method for controlling a process. The adaptive controller of the present invention monitors the output of the process. The method of the present invention also automatically controls the process optimally by tuning the parameters of the adaptive controller. Process control devices include devices (industrial controllers) that control process variables of a process having a predetermined transfer function and predetermined process parameters associated with the transfer function. The device has a first input terminal and a second input terminal, the first
a difference device that outputs a difference between a first signal coupled to an input terminal of the device and a second signal coupled to a second input terminal of the device. Also included is an adaptive controller connected to the difference device and process. The adaptive controller receives the error signal from the difference device and generates an output control signal that is coupled to the process. The adaptive controller is tuned to the process such that the output control signal causes the process to output a process variable signal with predetermined characteristics. The process variable signal is coupled to a second input terminal of the difference device and to the adaptive controller, thereby allowing the adaptive controller to monitor the process variable signal independently of the error signal. This configuration allows any change in the process variable signal as a result of a perturbation of the process parameter to be detected and adjustments to the tuning of the adaptive controller are made at that time.

A method of adaptively tuning an adaptive controller includes (on receiving an input instructing to start a control process) a step function for an output control signal having a value that equalizes a process variable signal to a set signal value. This includes the process of outputting. The set point signal is input to the process control device. Process variables are detected to identify the process and tuning parameters for the adaptive controller are calculated. After lineout occurs, process variable signals are monitored. If a change occurs in the process variable signal without causing a corresponding change in the set point signal, the tuning parameters are adjusted to compensate for the change in the process parameter. In other cases, the error signal is monitored if no change occurs in the process variable signal. When a new set point value is entered, the method begins by outputting a step function to identify the process, and the method is iterated.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Reference is first made to FIG. 1 in which a block diagram of a control device including an adaptive controller of the present invention is shown. Control device 1
includes process 10. This process involves temperature, pressure,
It outputs a signal of a predetermined parameter, such as a flow rate, a process variable (pv). The input signal to the control device 1 is a value indicating a desired value of an output or a process variable at a set point (sp). The set point signal and process variable signal are input to difference device 23. The difference device outputs an error signal e. This error signal is coupled to tuner 21 and controller 22. Controller 22 may be a well-known microprocessor. Tuner 21 and controller 22 operate in conjunction with each other to constitute the adaptive controller of the present invention. Controller 22 receives the error signal and produces an output signal Co. This output signal Co is the output signal of the process 10 sp=p
The signal is made to follow the input signal so that the signal is set to v. A lineout is said to have occurred when sp=pv. Difference device 2
3, the tuner 21, and the controller 22 can be packaged together to form the industrial controller 20. In a preferred embodiment of the invention, the output signal (process variable pv) is directly coupled to controller 22. In this way, the controller, under software control, cuts off the input of the error signal and
By inputting a process variable signal, the process 10
can be directly monitored. In this way, adaptive controllers 21 and 22 (sometimes referred to herein as controller 22 for simplicity) initially identify processes (described below) and tune themselves. If a change in process parameters occurs, it can be easily detected by the adaptive controller. The detected change may be one or more of process gain, time constant, or dead time (ie, a change in process 10). Changes in the process 10, and therefore in the process parameters, may cause pv to oscillate and cause pv to damped oscillations. In that case, in order to maintain pv at the set point, controller 22 seeks compensatory changes in the PID parameters. Thus, the adaptive controller of the present invention automatically "retunes" itself in response to changes detected in process 10 without making a corresponding change in the setpoint signal.

Reference is now made to FIG. 2A, where a block diagram of a typical controller is shown. The control device shown in FIG. 2A includes a device 200 having an input terminal, an output terminal and a transfer function T(s). The input signal IN and the output (OUT) of device 200 are connected to difference device 23 . This difference device outputs the error e to the input terminal of the device 200 such that OUT=IN[T(s)/{1+T(s)}]. In response to the step input, OUT=IN[T(s)/S{1+T(s)}].

Reference is now made to FIG. 2B, which shows a block diagram of an apparatus similar to apparatus 200 shown in FIG. 2A of the controller. In this figure, the device 200 of FIG. 2A is divided into two devices 201 and 202. Devices 201 and 20
2 is equal to the transfer function of device 200 of FIG. 2A. Therefore, the transfer function T(s) of the divided device of FIG. 2B is T(s)=G1(s)·G2(s).

[0011] Here, pv=process variable (actual value)
sp = set point (desired or commanded value); G1(s) = transfer function of industrial controller 20 (or controller 22); G2(s) = transfer function of process 10.

Most industrial processes consist of one or two delays with or without downtime. Therefore, each transfer function takes the form: a) Single lag without dead time G2(s) = K/(s+a) b) G2(s) for double lag without dead time
= K/(s+a)(s+b)c) G2(s)=Ke-DTs/(s
+a) d) G2 for double delay with dead time
(s)=Ke-DTs/(s+a)(s+b)

001
3] The transfer function G2(s) of the controller 202 is the process 1
Since it is a transfer function of 0, it is essentially given. Therefore, control over process parameters is not possible. The transfer function G1(s) of the device 201 is the transfer function of the controller 20. This transfer function G1(s) can be controlled, and G1(s)=Gc(STi+1)(sTd
+1)/{sTi(asTd+1)}.

K(s)=Gc (gain) I(s)=
The controller 201 is further divided as shown in FIG. 2C, such that (Tis+1)/Tis S(s)=(Tds+1)/(aTds+1).

Therefore, G1(s)=K(s)I(s
)S(s), and for this process G2(
s) are of the following types: a) Single lag G1(s) with no dead time = K(s) I(s) b) Double lag G1(s) with no dead time = K(s) I(s) S(s) c) Single lag G1(s) with dead time = K(s) I(s) d) Double lag G1(s) with dead time = K(s) I(s) S(s)

[0016]
Depending on the process 10 (i.e., G2(s)), the controller 20 sets the appropriate constant to some predetermined value by switching at least one of K(s), I(s), and S(s). or to zero or any other equivalent means) to produce a marginally damped system. The tuning constants (or PID constants corresponding to K, I, and S) to be determined are: Gc = Controller gain Ti = Integral time constant Td = Differential time constant a = Rate gain coefficient. In the preferred embodiment of the invention, 0.1
It is fixed at 25. It is.

Reference is now made to FIG. 3, where a flow diagram of the adaptive control method of the present invention is shown. The method begins by initiating a control process (block 300). In the startup process, the user brings the process to about 10 percent of its scope in manual mode. The user selects a set point for a low value process variable. The process variable is typically about 10% of full power performance (for example, if the process variable is to control a valve, the process variable will command about 10% of the valve to open, and the process variable will control the furnace. process variables such as turning on the burner 10% to obtain a lower heat output). Finally, when a lineout occurs (sp
= pv) In this start mode the process is almost in standby mode.

[0018] When starting automatic control, the desired set point is input to the controller and signaled to begin the automatic process. The announcement is a manual action such as pressing an "auto start" push button (block 305).
). Control 20 calculates the output value Co. This output value predicts that the process variable is equal to the set point. This calculation is based on the required controller output to place pv at its 10 percent position. Controller 20 outputs a step signal to the calculated value (block 310
), begins identifying dead time (block 315). The controller looks for the process variable to start moving after the output of the step change and measures the time it takes from the step change to some predetermined small increase in pv value. This time is called the dead time DT. If the process variable value increases immediately upon outputting the step signal, the controller determines that there is no dead time (DT=0). The method continues with the identification of the type of process 10 contained in the control device 1 (block 320). When the pv value begins to rise, controller 20 measures the slope. Assuming that the slope decreases continuously from the start of the rise,
The process is identified as a single lag. The process time constant T1 and steady-state gain K are calculated from the following two equations. T1=(PV2-PV1)/(PV1'-PV2')K
=(PV2+PV2'T1)/Co Here, PV1 and PV2 are the PVs during their rise.
These are the two measured values. PV1' and PV2' are two slope measurements corresponding to PV1 and PV2, respectively. Co is the step size of the controller output.

[0019] When the PV value begins to rise, the controller measures the slope. As the slope rises to a maximum value and then falls, the process is identified as having two lags. The time t1 required from the start of the ascent until reaching the point of the highest slope is measured. The process time constants T1, T2 and steady state gain K are calculated according to the following procedure. Reference is now made to FIG. 4, where the curves of process variables as a function of time for a two-lag system are shown.

From the general formula KCo=PV+(T1+T2)PV CoK=mfnt1+PV1 is easily guided. Here, m=tanx=PV1' PV1=process variable at t1 It is. Furthermore, N fn 10 4.3 8 3.8 6 3.3 4 2.7 2 2.2 Here, n=T1/T2 fn=(T1+T2)/t1 It is.

[0021] Measure t1, PV1, PV1', and
By knowing and selecting fn=3.3, K, T1
, T2 can be reasonably approximated. This approximation is Gc,Ti
, Td (block 325).

As the process approaches lineout (block 330), the exact value of K is measured and fn
A new value of is calculated. From this new values of T1 and T2 are calculated and the controller is retuned more accurately (block 335).

After the retuning shown in block 335, the controller monitors pv and sp. If the process parameter changes, but then without a corresponding change in set point sp (block 340), a change in the process parameter is determined. p without corresponding change
Assuming no change in v occurs, the set point is monitored to determine if a new set point is entered (block 345). Assuming no new set point is entered, controller 22 continues to monitor the system to maintain the desired output pv. If a new set point is entered (block 345), the process begins at the output step of block 310 and repeats dead time identification and process identification in calculating process parameters (blocks 310-335).

If a change in the process variable is detected without causing a corresponding change in the set point (block 340)
, changes in process parameters are determined.

Reference is now made to FIG. 5, where a flowchart of process change compensation is shown. If the PV is caused to oscillate at a frequency wo by changing a process parameter (block 400), then wo≦wi, where wi=1/Ti' (Ti is the integration time constant). Controller 22 shifts the notch so that wi=0.5wo. If wo>wi, the controller shifts its notch so that wd=wo. Here, wd=1/Td (
Td = differential (rate) time constant). If the oscillation persists, the controller cuts its gain in half.

Assuming that changes in the process cause the PV to damped oscillations at frequency wo (block 405),
The controller 22 sets wd=wo. If a process disturbance occurs and the time t to return to lineout is t>DT+T1+T2, controller 22 shifts the frequency notch higher by multiplying wi and wd by 1.3. Here, t is time DT=non-operating time T1, T2=process time constant.

If the process gain changes such that a different controller output is required to maintain PV at the same set point (block 415), then the new process gain is determined by: Kn=Ko(Coo/Con) where Kn=new process gain Ko=old process gain Coo=old controller output Con=new controller output.

As described above, the process controller automatically adapts to any changes in the process without external intervention.

Controller tuning is provided in Appendix A. 3
Other processes, such as process 10 having two delays, can be identified in the preferred embodiment of the present invention. These processes typically manifest themselves as two delays and a dead time. A preferred embodiment of the invention identifies such a process,
Controller 22 is appropriately tuned as described above.

Appendix A Generally, three controller tuning methods are used. The methods apply to set point changes in processes with and without downtime, and to process parameter changes in any process.

The transfer function of the controller is a PID whose real number is zero.
It is an equation. C(s)=Gc(STi+1)(STd+
1)/STi(SaTd+1) (
A) where, Gc=controller gain Ti=integral time constant Td=differential time constant a=rate gain coefficient (a=0.125).

For set point changes in a process without dead time, the controller tuning method is polar cancellation. The zero in equation (A) is tuned to cancel the process pole in equation (B). P(s)=K/(ST1+1)(ST2+
1) (B) Here, K=
Steady state gains of the process T1, T2=process gain (T1>T2).

In polar cancellation tuning,
2 delayed processes
One delay process Ti=T1

Ti=0.16T1 Td=T2
T
2=0 Gc=6/K
Td=0

Gc=24/K.

For set point changes in processes with dead time, the controller tuning method
This method was developed by E.B. Dahlin. The process transfer function is P(s)=Ke-DTs/(ST1+1)(ST2+1
). Here, DT=dead time.

[0035] In darlink tuning,
2 delayed processes
One delay process T1=T
i+T2
Ti=T1 Td=T1T2/(T1+
T2) Td=0 G
c=3/K(1+3DT/T2) (same as the equation on the left).

[Brief explanation of drawings]

FIG. 1 shows a block diagram of a control device including an adaptive controller of the invention that utilizes the method of the invention.

FIG. 2 shows a block diagram of a portion of a typical different controller.

FIG. 3 shows a flowchart of the adaptive control method of the preferred embodiment of the present invention.

FIG. 4 shows typical curves of process variables as a function of time for a two-lag process.

FIG. 5 shows a flowchart of process variation compensation in a preferred embodiment of the present invention.

[Explanation of symbols]

21 Tuner 22 Controller 23 Difference device

Claims (2)

[Claims] 1. A device (industrial controller) for controlling process variables of a process having a predetermined transfer function and predetermined process parameters associated therewith, having the following a) and b): a) having a first input terminal and a second input terminal, the difference between a first signal input to the first input terminal and a second signal input to the second input terminal; Difference means to output. b) connecting the difference means to the process, receiving an error signal from the difference means and generating an output control signal coupled to the process, the output control signal processing a process variable signal having predetermined characteristics; adaptive controller means tuned to cause an output of the process variable signal to be input to a second input terminal of the difference means and to the adaptive controller means, whereby the process variable signal is input to a second input terminal of the difference means and to the adaptive controller means; Adaptive controller means for making tuning adjustments by monitoring the process variable signal independently of the error signal to detect changes in the process variable signal. 2. An adaptive controller for controlling a process variable of a process having a predetermined transfer function and a predetermined process parameter associated with the transfer function, the adaptive controller being connected to the process. starting a control process in a method for tuning an adaptive controller in the process such that the output control signal causes the process to output a process variable signal (corresponding to a value associated with the process variable) having predetermined characteristics; a) outputting a step function for the output control signal having a value such that the process variable signal equals the value of the set point signal input to the process controller; ,
b) detecting process variables to identify the process; c) calculating tuning parameters for the adaptive controller; d) monitoring process variable signals after lineout has occurred; e) assuming that some change occurs in the process variable signal without causing a corresponding change in the setpoint signal, i) adjusting the tuning parameter to compensate for the change in the process parameter; and ii
) continue to step (f) and otherwise, if no change occurs in the process variable signal, iii) step (f)
f) Given a given set point value is input, i
A method for adaptively tuning an adaptive controller, characterized in that the method comprises the steps of: ) repeating the step in which step (a) begins, and otherwise ii) repeating step (e).